Do you trust your eyes or the numbers?

[jammer_jammer]

I've been reading books on post processing and I think I may have fallen into

the trap of trusting the numbers more than my eyes. For instance, I'm now

scared to death to let any part of the image approach 250-255. In order to

achieve this with some of my very contrasty, scanned, Kodachromes, it is at the

expense of the over all image because allowing no blowouts anywhere is

sometimes at the expense of having an image that is just too dark to enjoy. I

had gotten so used to this that I guess I was to the point of not really

noticing how dark my images really were but it has been brought to "light"

recently that my images are too dark, so I'm rethinking the whole thing.<P>

 

So,....today I started looking at others images that I really love and noticed

that there are sometimes quite a few areas that are completely blown out and

yet it doesn't seem to hurt the image. As long as the main focus of the photo

is not blown and the areas that are blown are not too destracting, is there

anything wrong with improving the majority of the image by allowing some areas

to blow out?

[godfrey]

The numbers are sometimes important, but what the photo looks like is always more

important. Trust your eyes.

 

Godfrey

[phoneguy]

The numbers are secondary to what is pleasing to the eye. (IMHO)

[roger_smith4]

Well, if the scan out of the scanner is clipping white and blacks, try merging multiple exposures in a program like Photomatix (not the HDR feature!).

 

Consider Photoshop Lightroom which lets you "recover highlights" AND increase exposure of the rest of the image. It's quite impressive.

[richard_cochran]

The rule of thumb I was taught, from long before the digital era, was that every photo should have some absolute pure white and some absolute pure black, if only a tiny area of each. A goal was to get a "full scale print" of your negative. I can't remember if I read that in some of Ansel Adams' works or not, but it was commonly repeated wisdom, whatever the original source.

<p>

Like all photographic rules of thumb, adherence is optional and occasional informed rule-breaking is encouraged. But definitely don't avoid ends of the histogram just for the sake of avoiding them.

[qtluong]

<i>is there anything wrong with improving the majority of the image by allowing some areas to blow out?</i>

<p>

It's OK if you can be sure that those areas are unimportant, and that they do not become a distraction. However, if you learn to use masking tools such as layers, you will be able to make local corrections that do not affect certain areas.

[mendel_leisk]

Using a modified Vuescan raw file worklfow, my first gamma 1.0 scan output usually has little if any blow-out. Subsequently, I will produce a finished file from that, through Vuescan's scan-from-disk. My start point with this process is to clip (in Vuescan's Color Tab) 0.5%. For most slides this works fine. If there is important detail in the highlights, I'll occasionally reduce the clip to a smaller amount, or zero.

[j_sevigny]

Your second paragraph nails it for me. Photoshop geeks get all bent out of shape by blown highlights, and in some cases, there's good reason for it. On paper, blown out highlights look like paper and nothing else.

 

But a lot of photojournalistic work shows blown out skies for the obvious reason that the shooter is opening a few stops to show what's happening in front of him/her.

[Tim_Lookingbill]

Blow outs are only used to compensate the dulled luminance of

the paper it's printed on. Same goes for scanning film. If the

actual gelatin like feel of the slide becomes too much part of the

image's density range then blow outs will have to be applied

because you want scene correctness not the texture of the film to

be a part of it.

 

I used to operate a graphics camera for an entertainment tabloid

printed on dull, dark yellowish newsprint. To make the B&W

images pop we had to overexpose to induce a percieved

increase in contrast. Blowing out the dots on the highlites of

people's faces was necessary. But if I had those same line

conversion dot screens printed on slick, bright glossy stock the

images would look like crap.

 

But the dark/blue appearance as seen in Jammer's scans in

another thread is from a lack of proper gamma encoding when

rendering the scene because most scanner device's capture

data linearly. You have to apply the proper descriptive profile that

describes this linearity and assign it to the image or increase the

exposure through the scanner's hardware like shining a brighter

light bulb through the film so as to blow out the clear portion of

the film to 255 white. Usually you end up losing quite a bit of

detail doing it this way.

 

My minilab scans add the big bold S-curve high contrast

oversaturated treatment to my negative because its scanner is

set to be balanced for the printer's dull white paper. Just

because the scans are written to sRGB still doesn't make it right.

[robert lee]

"I'm now scared to death to let any part of the image approach 250-255."

 

This just needs to be elaborated on. The problem is not with parts of the final image being blown out (think about the specular highlights from the surf in the cliff and beach image you posted previously.) The problem is with clipping the intermediate layers while operating on the image.

 

Say you've applied a filter that pushes a channel of some pixels above 255. Once that image information is lost, no application of any subsequent filters can bring it back. The final image will have uncorrectable artifacts.

 

By the way, this is one reason why it's desirable to edit with the most number of bits per pixel the imaging software will allow.

[danny_wong2]

Unless I know I can't trust them due to medical evidence, I will alwasy trust my eyes even with vision correction lens.

[micheleberti]

<i><blockquote>[...]In order to achieve this with some of my very contrasty, scanned, Kodachromes, it is at the expense of the over all image because allowing no blowouts anywhere is sometimes at the expense of having an image that is just too dark to enjoy.</i></blockquote>

 

Have you ever tried using masks in conjunction with contrast enhancements techniques? Usually this gives you extra degrees of freedom .

[jammer_jammer]

Thanks everybody!<p>

 

Tim,<br>

After a lot of experiment, my approach with dark or very contrasty slides is to set the gain so that the highest high is around 250 and do the rest in PS. If that setting leaves me with almost no information in the mid range and shadows, then I'll do another scan and expose for the shadows and then merge the two images in PS.<p>

 

I'm so happy I could just spit about the fact that the film that I chose to take thousands of images with, ends up being the very MOST difficult film to drag a decent digital image out of. It also doesn't help matters that at the time I was addicted to underexposing for to get that nice color saturation.: -(<p>

 

I guess the real important thing is to not get rid of the original scan/scans. That way, I can always go back to square one if need be.<P>

 

Boy, I can't WAIT to have this scanning project over with. I then plan to move onto my first DSLR. I know that will have it's learning curve and challenges also but after this never ending project is over, NO MORE SCANNING FOR THIS GUY. : -)

[Tim_Lookingbill]

Jammer,

 

I feel for you. I can tell you I had trouble with your beach image so

I know what you must be going through on the rest. Since you

didn't indicate how close I got with your original slide, I assumed

it solved your problem.

 

Your scans look that way because it's in machine color straight

off the scanners sensors without any gamma encoding or idea

what color is suppose to look like because your display shows

color differently with P22 phosphors(LCD mimic this) and a

gamma correction which the scanner sensors/software knows

nothing about so the numbers are pretty much meaningless.

There's know adjustment to any vLUT like Photoshop does when

reading your display profile in tagged images.

 

My Agfa Arcus scanner produced the same looking scans from

prints back in '98 before I knew about color management. I spent

hours fixing the color like you while teaching myself color editing

at the same time. I found out later the color management portion

of the scanner software was broke.

 

As soon as PS 5 came out and color management took off, Agfa

upgraded their scanner software with better CM integration my

Mac system which included a Source, Display and Destination

CM dialog box where I could choose either Kodak, Fuji or Agfa

Reflective media scanner profile, my custom monitor profile as

Display and a working space of choice as Destination and I was

done. An incredible perfect match and no more dark and blue

scans.

 

Your Nikon should have the same thing. Otherwise you'll need to

induce your own gamma encoded appearance to the image

editing in Photoshop and saving these settings and apply to the

rest of your images.

 

I would suggest you start with a big move with the gamma slider

first like around 1.50 if the other images look as dark as the

beach scene after establishing 255 spectral highlite. Then work

on fixing the color temp starting with the blue middle slider.

Afterward apply an s-curve to brighten the contrast.

 

Or upgrade your Nikon scanner software or use Silverfast or

Vuescan.

[eric merrill]

Both.

 

Once you start looking at an image too long, your eyes are no longer reliable. Use the numbers as sanity checks.

 

 

Eric

[jammer_jammer]

Tim and Eric,

Thanks<p>

 

Tim,

The Nikon V ED has a choice between negative, postive and Kodachrome postive. Believe it or not, I am using the Kodachrome setting. Can't imagine what I'd have if I didn't. <p>

 

One thing I started doing today that is saving me quite a bit of time, is to work up all my curve, levels, and color balance layers for a particular image. I leave those layers intact and then for all the subsequent images taken the same day under the same light, I just drag all those previously made layers from the first image, into the next image. Of course there is still a little tweeking but at least I'm not starting from scratch on each file.<p>

 

Thanks again Tim